Rényi-type empirical processes (Q1192007)

The paper develops a theory of asymptotics for Rényi-type weighted empirical and quantile processes and statistics. Let \(\alpha_ n(t)\) be the uniform empirical process, \(\tilde \alpha(t)=\alpha_ n(t)\text{\textbf{1}}_{[0,1]}(t)\), \(q(t)\) a weight function, \(\tilde q(t)=t\text{\textbf{1}}_{[0,1]}(t)+tq(1/t)\text{\textbf{1}}_{(1,\infty)}(t)\) and \(\{a_ n\}\) a sequence of positive numbers tending to zero. In the case of moderate weight functions the processes \(a_ n^{-1/2}\alpha_ n(a_ nt)/q(t)\) and \(a_ n^{-1/2}\alpha_ n(1-a_ nt)/q(t)\) are approximated in sup-norm with a sequence of weighted Wiener processes \(\{W_ n(t)/q(t)\}\). However, heavily weighted processes \(n^{1/2}\tilde\alpha_ n(t/n)/\tilde q(t)\) and \(n^{1/2}\tilde\alpha_ n(1-t/n)/\tilde q(t)\) are approximated in sup- norm with a sequence of weighted Poisson processes. These are approximations of tail empirical processes. In the case of heavy weights, an analogous approximation of the middle section of \(\alpha_ n\) is also obtained. All results are extended to uniform quantile processes.